- Email: [email protected]
Visualization of place attachment
Applied Geography 99 (2018) 77–88
Contents lists available at ScienceDirect
Applied Geography journal homepage: www.elsevier.com/locate/apgeog
Visualization of place attachment a,∗
Brad Maguire , Brian Klinkenberg a b
T
b
Department of Geography, Vancouver Island University, 900 Fifth Street, Nanaimo, BC, V9R 5S5, Canada Department of Geography, University of British Columbia, 1984 West Mall, Vancouver, BC, V6T 1Z2, Canada
A R T I C LE I N FO
A B S T R A C T
Keywords: GIS Mapping Place Place attachment Sense of place Visualization
The concept of place attachment is amorphous and difficult to study. To address this, we set out to map and analyze place attachment for a city park in Nanaimo, British Columbia, Canada. We developed a GIS application called the Place Analysis System (PAS) to allow place attachment data to be collected, georegistered, stored and processed into map form. Upon analysis, we found notable differences in the spatial distribution of place attachment between groups and under different weather conditions. The ability to visualize and analyze place attachment can assist with both academic and applied studies in fields such as planning.
1. Introduction Recent research into place attachment has revealed the major components of this concept. Nevertheless, place attachment remains ethereal, lacking in form and context. Despite extensive study in the geographic literature, little attention has been given to the spatial distribution of place attachment. We addressed this problem by connecting place attachment to realworld places. Using a geographic information system (GIS), we mapped the strength and spatial distribution of place attachment around the features that make up places. This makes place attachment easier to visualize and measure, making it more amenable to empirical analysis and applied research. 1.1. Understanding place Place studies can be complicated because there are many concepts of place from different fields of study and intellectual traditions. Williams (2014) defines three views of place: demos, which views place as being dynamic, cosmopolitan, and spatiotemporal, ethnos, in which place is defined by group identities and parochial attachments to place, and bios, in which place is defined by bioregionalism and a “back to the land” ethic. Our research examines place primarily from the demos perspective. Many of the concepts used to study place benefit from the work of Massey (1997), Pred (1984) and Thrift (1994). This perspective re-
∗
cognizes the cosmopolitan nature of place and leads us to model place as unbounded and connected to other places. The dynamic nature of place can be analyzed by studying where people travel at different times and under different weather conditions. 1.2. Modeling place using GIS In this paper, we introduce and demonstrate the use of the Place Analysis System (PAS), a GIS application that facilitates improved data collection and analysis of place based on place attachment theory (Stedman, 2002; Williams & Vaske, 2003). The PAS enables researchers to create fuzzy models of place attachment, which facilitate the understanding, analysis and communication of this concept. We will demonstrate how the system can help visualize place attachment, taking study participants’ vague notions of place attachment and making them concrete. 1.3. Study area We chose Colliery Dam Park, a 27.2 ha (67.2 ac.) semi-wilderness park in Nanaimo, Canada for our study area (Fig. 1). The centerpieces of the park are two artificial lakes formed by the Colliery Dams (Fig. 2). These lakes formerly supplied water for coal mining, but now serve recreational purposes (Fig. 3). The park has much in common with larger parks such as Central Park in New York, Rock Creek Park in Washington, D.C. or Stanley Park in Vancouver.
Corresponding author. E-mail address: [email protected] (B. Maguire).
https://doi.org/10.1016/j.apgeog.2018.07.007 Received 30 August 2017; Received in revised form 28 June 2018; Accepted 2 July 2018 0143-6228/ © 2018 Elsevier Ltd. All rights reserved.
Applied Geography 99 (2018) 77–88
B. Maguire, B. Klinkenberg
now moving towards decentralized or mobile computer data entry through web pages (Carver et al., 2009), and smartphone applications. In these GIS studies, places have been conceptualized in different ways. They have been represented as points (Brown & Raymond, 2007; Brown, 2005; Brown et al., 2014), polygons (Lowery & Morse, 2013), or rasters. Carver et al. (2009) developed a method that uses a “spray paint” metaphor, in which a participant can “paint” place importance onto a raster map. This technique allows for the direct recording of a fuzzy surface.
Abbreviations PAS
Place Attachment System, the GIS application used to collect and organize data for this study
2. Literature review 2.1. Working with fuzzy surfaces
2.4. Research advances
Individuals view places differently from one another (Williams, 2014) and have different features and levels of interest. We can represent these as fuzzy surfaces (Burrough, 1996; Fonte & Lodwick, 2005; Fonte, 2008; Zadeh, 1965), in which place attachment decays as distance increases. This decay represents the decrease in awareness as features fade from short-term memory. The awareness never reaches zero because long-term memories of a feature may be recalled at any time with the right stimuli.
The method proposed in this paper offers a general solution for the modeling of place attachment based on the demos concept of place. Collecting the data soon after participants have visited the features allows us to work with participants while impressions of new features are fresh. In addition, the sights, sounds, and smells of the park assist in the recollection of past memories of park features. As well, participants can use body language to indicate the distance, direction, size and configuration of the features that they are describing. Brown (2005) mentioned that “… in the Chugach National Forest study, many of the intrinsic value dots were placed on empty spaces …” (p. 33). Although Brown attributes this to people recognizing the intrinsic value of wild places, another possibility is that a lack of familiarity with certain places or a lack of map literacy caused people to place the dots inaccurately. Brown also relates that the points were used to infer polygon boundaries based on the density of their placement, which is a subjective process. Written and verbal communications were used when interacting with the participants to avoid issues related to computer literacy. Because we mapped the features in advance, we were able to use placebased nomenclature (Evans & Waters, 2007; cited in Carver et al., 2009) to match the names of features with their mapped counterparts. This allowed features to be identified without having to rely on the cartographic or map reading skills of the participants. Once identified, the features were automatically copied for analysis so that the original data remained intact. The decay in place attachment was calculated for each feature, and this was used to create feature surfaces that describe how memories of features decrease as distance increases. Rather than using a pre-defined interpolation method (e.g. spline, inverse distance weighting), the shape of each feature surface was based on the decay rate of place memories, as described by Dornič
2.2. Modeling emotion We employed Plutchik's (1962, 1980) psychoevolutionary model of the emotions as the dictionary of emotions used in this study. A numerical intensity value is associated with each emotion in Plutchik's model. This value is used in the PAS; the emotion name simply provides a convenient reference for the participants to use. 2.3. Efforts to map place Geographers and psychologists have been refining the concept of place over the past six decades. Work on “mental maps” has revealed the distortions present in how people model their surroundings (Gould & White, 1992). These distortions make it difficult to operationalize findings, because it is difficult to identify exactly where features are located (Golledge, 1978, p. 41). The work of Lynch (1960) has been built upon and expanded by modern researchers using GIS. There has been a gradual shift from paper-based to computer-based methods of recording spatial data. Early work, such as Brown (2005) made use of paper-based methods for data collection, whereas later work (Brown & Raymond, 2007; Brown, Raymond, & Corcoran, 2014) made use of computer data entry. We are
Fig. 1. The study area around Colliery Dam Park.
78
Applied Geography 99 (2018) 77–88
B. Maguire, B. Klinkenberg
Fig. 2. Colliery Dam Park and the study area.
Fig. 3. The lower lake and dam at Colliery Dam Park (photo by B. Maguire).
submissions were skipped. Out of 302 responses, 63 were rejected, leaving 239 responses for analysis (Table 1). A Kruskal-Wallis H test was used to evaluate whether the three different sampling methods yielded similar place attachment values. An examination of boxplots showed similar distributions in the place attachment values, which was confirmed by test results that showed similar overall place attachment values (H(2) = 3.091 p = .213). Based on these findings, we are confident that all three sampling methods can be combined to create an overall composite view of place attachment for Colliery Dam Park.
(1967). We term this a theory-based interpolation method. The software was designed to combine feature surfaces for related objects. For example, if a participant describes “water features” we can combine all the lakes (polygons) and rivers (lines) to create a single composite feature. From this, we construct a composite surface, which is simply the combination of all component feature surfaces. In turn, the composite surfaces for all features are combined to create a single place attachment surface for each participant. 3. Data and methods 3.1. Sampling strategy
Table 1 Numbers of participants by sampling method and rejection status.
We collected data from six sampling points at well-traveled locations from Oct. 20, 2011 to Sept. 29, 2012 (Fig. 2). The day of the week (including weekends) and time of collection (morning, afternoon, or evening) were rotated to maximize the chances of interviewing people from different sub-populations of users. Samples were mostly collected randomly, except at times of low traffic, when convenience sampling was employed. A considerable number of park users volunteered to take the survey. In cases where participants provided invalid information or demonstrated that they did not understand the questions, their
Not Rejected Rejected Total
79
Randomly Sampled
Volunteered
Convenience Sampled
Total
142 (47.0%) 32 (10.6%) 174 (57.6%)
85 (28.1%) 26 (8.6%) 111 (36.8%)
12 (4.0%) 5 (1.7%) 17 (5.6%)
239 (79.1%) 63 (20.9%) 302 (100.0%)
Applied Geography 99 (2018) 77–88
B. Maguire, B. Klinkenberg
Fig. 4. Main survey questionnaire as filled out by participant 362199345.
To help participants complete the questionnaire, they were provided with a double-sided reference card on which Awareness Distance was explained (Fig. 5). This card also listed 53 emotions in addition to “no emotion.” Thirty-two of these were shown on Plutchik's wheel of emotions (Plutchik, 1962), and those remaining were shown in a table. If a participant mentioned an emotion that was not described on the reference card, the software allowed any of the 118 emotions described by Plutchik (1962) to be chosen. For each feature, participants were invited to provide qualitative information including descriptions of the feature, photographs, sketch maps, or explanations of why it was important. These sources of information provided context for the place attachment surfaces produced by the PAS. Participants were also asked to provide their age, gender, ethnicity, occupation, level of education, and postal code. This demographic information showed that study participants were divided roughly equally between genders and age groups, except for minors (under age 19) who required parental assent before they could participate (Table 2). Based on the self-reporting of educational status, 55 participants (18.2%) had an elementary education, 39 (12.9%) had a secondary education, 165 (54.6%) were university or college graduates, and 43 (14.2%) had post-secondary degrees. On the final page, the survey taker noted the date, time, weather conditions and location of data collection. The time of collection is associated with all data, so that spatiotemporal changes to place attachment can be examined.
3.2. Data collection Participants were asked to fill out a paper questionnaire to identify an open-ended number of park features (Fig. 4). After providing the feature name, they were asked to indicate the main emotion associated with the feature, which was used to determine the emotional intensity. Next, participants rated the importance of the feature on a 7-point Likert scale (1 = very unimportant, 7 = very important). The importance was used as an indicator of place dependence and the intensity of the emotion was used as an indicator of place identity. Participants next indicated the distance at which they were no longer aware of the feature. The awareness distance is defined as “How close you must be before you become aware of it [a feature] when approaching, or how far away you must be before you lose awareness of it when you are leaving.” Awareness distance (the XY value) measures the Euclidean distance from the nearest component of a composite feature and is used to calculate the decay of place attachment (the Z value) as distance increases. It extends beyond the visible range to where the feature is no longer in short-term memory. For first-time visitors to a place, awareness distance is the same as visual distance, but with repeated visits, place attachment increases, knowledge of the place grows, and a person's mental map becomes more complete (Gould & White, 1992; Kitchin & Blades, 2002). Through this process, people gain configurational knowledge of important features that are not visible from particular locations. The awareness distance is uncorrelated with the size of the feature. A single point feature can have a larger awareness distance than a large polygon feature. It should be noted that in such a case, the total area within the awareness distance of the polygon feature may be greater than that within the awareness distance of the point, simply because the polygon has a larger area than the point.
3.3. Data entry A researcher entered data from the paper form into the PAS. The PAS interface (Fig. 6) mimics the paper form to facilitate data entry and reduce transcription errors. Validity checks are performed (e.g., 80
Applied Geography 99 (2018) 77–88
B. Maguire, B. Klinkenberg
Fig. 5. The reference card that was provided to participants to assist them with filling in the survey.
Table 2 Breakdown of study participants by age and gender. Age
Unreported
< 19
19–29
30–39
40–49
50–59
> 60
Total
Males
22 (7.3%) 9 (3.0%) 31 (10.3%)
4 (1.3%) 5 (1.7%) 9 (3.0%)
33 (10.9%) 29 (9.6%) 62 (20.5%)
21 (7.0%) 18 (6.0%) 39 (12.9%)
19 (6.3%) 26 (8.6%) 45 (14.9%)
24 (7.9%) 36 (11.9%) 60 (19.9%)
31 (10.3%) 25 (8.3%) 56 (18.5%)
154 (51.0%) 148 (49.0%) 302 (100.0%)
Females Total
method used. All three tables use a shared 9-digit random key to allow the tables to be joined for analysis, while preserving the anonymity of study participants.
incorrect postal codes are rejected) to help ensure that data are entered correctly. To enable the use of place-based nomenclature, each named feature is matched with one or more previously mapped features. The software presents a list of 1322 mapped features to choose from as well as a series of buttons to help the researcher choose composite features (Fig. 7).
3.5. Data processing Once the data have been stored in the database, they can be processed to build a place attachment surface for each participant. For each feature, a feature surface is created as a floating-point raster with a 2.5 m (8.2 ft) resolution, which maintains detail, but still allows for efficient data processing. Although the centering of the point, line, and polygon features is exact, the curvature of the feature surfaces is theoretical, based on the work of Dornič (1967) (Equation (2)). The feature surfaces are then combined using a Fuzzy OR function (Zadeh, 1965) to create a composite surface for each composite feature (Fig. 8). When feature surfaces overlap, the composite surface maintains the highest input value.
3.4. Organization of data Once the data have been entered and validated, the system automatically transfers it to a relational database for analysis. Three main tables are used in the database. Mainform_Features stores information about the features described. Mainform_Participant stores information about the participants, such as frequency of park visits and demographic data. Mainform_Questionnaire stores the time, date and location of data collection, the weather conditions, and the sampling 81
Applied Geography 99 (2018) 77–88
B. Maguire, B. Klinkenberg
Fig. 6. An analog of the paper form reduces the likelihood of transcription errors.
All composite features are then combined using a Fuzzy OR function to create a place attachment surface for the participant. Fig. 9 provides an example of how a place attachment surface is built. A planimetric view of this place attachment surface is shown in Fig. 10a. The place attachment surface is scaled to a range of [0.0, 1.0] so
We next scale the Z-value of the composite surface [0.0, 1.0] by the place attachment value for each feature (Equation (4)). The scaled composite surface ranges from the place attachment value at the feature outward to the minimum place attachment value (1.7 …) at and beyond the awareness distance.
Fig. 7. The Choose Feature window allows mapped features to be assigned to each feature listed by a participant. The buttons help to quickly select individual or composite features. 82
Applied Geography 99 (2018) 77–88
B. Maguire, B. Klinkenberg
Fig. 8. Composite surfaces for point, line and polygon features, which are shown in black on top of each surface. The lower line for each surface shows the awareness distance. A composite feature may be composed of multiple elements, such as the boardwalk in the center image. Features may also have holes, such as those in the polygon on the right.
Where:
that it can be compared and combined with other participants’ results (Equation (5)). The range [1.7 …,11.74] in Fig. 9 is scaled to [0.00, 0.86] because the place attachment surface has the lowest possible minimum value (1.7 …), but not the highest possible maximum value (13.32) . Because each fuzzy surface extends away from features that have been accurately mapped, the surfaces for different participants can be measured, compared and summed together. This allows a summed fuzzy surface to show the “consensus view” of a group or the entire population of study participants.
A is the place attachment value; I is the place identity; and D is the place dependence of the feature. Substituting the minimum and maximum place identity and place dependence values into Equation (1) produces an interval of [1.7 …, 13.32 ] for place attachment values. The curve of the decay surface is based on research about how memories of emotions decay with distance. Ekman and Bratfisch (1965) found that the emotions related to a place follow an inverse square law, but the exponent was later refined by Dornič (1967) to yield Equation (2):
4. Theory/calculation Plutchik's intensity values are based on empirical research, in which a group of research subjects was asked to rate the psychological impact of many emotions, from 1 (low) to 11 (high). Plutchik (1980) found that emotions with intensity values below 3 were indistinguishable from one another, so values below 3 were not used. Combining the [3,11] interval for place identity (intensity) with the [1,7] interval for place dependence, we come up with the following formula for place attachment, in which place identity (9 values) is separate, but equal in importance to place dependence (7 values):
A = 7/9 (I − 2) + D
Zij = dij−0.47
(2)
Where: Zij represents the value of the decay surface; and dij is the distance from the feature. Because the function is asymptotic to zero, it supports Massey's (1997) assertion that places have no boundaries since such a surface extends around the globe. Equation (3) creates a normalized
(1)
Fig. 9. Place attachment surface for participant 362199345 viewed from the northeast, showing the contributions made by individual composite surfaces. The place attachment value shown for each composite surface is A, and the awareness distance is d. The minimum place attachment value has been removed for clarity.
83
Applied Geography 99 (2018) 77–88
B. Maguire, B. Klinkenberg
Fig. 10. Place attachment surfaces for four participants. Darker (redder) values represent more important areas, whereas lighter (bluer) values represent less important ones.
dawareness is the awareness distance expressed by the study participant; and dmin is the minimum possible distance.
logarithmic surface with the correct curvature, but which has a value of 1.0 at the center and 0.0 at and beyond the distance specified by the participant.
if dij = 0 ⎫ ⎧1 ⎪ ⎪ dij−0.47 − d awareness−0.47 > ≤ if d 0 and d d Mij = ij ij awareness ⎬ ⎨ dmin−0.47 − d awareness−0.47 ⎪ ⎪0 if dij > dawareness ⎭ ⎩
To create a place attachment surface, the composite surface produced in Equation (3) is scaled to ensure that values range from the feature's place attachment value (A) at the inner edge of the feature surface to the minimum possible place attachment value (Amin) at and beyond the awareness distance (Equation (4)).
(3)
Where:
Aij = (A − Amin ) Mij + Amin
Mij is the multiplier value for the place attachment surface; dij is the Euclidean distance recorded in each pixel;
Where:
84
(4)
Applied Geography 99 (2018) 77–88
B. Maguire, B. Klinkenberg
be made in relation to places.” We can see graphical evidence of this in Fig. 10, which shows a selection of four place attachment surfaces constructed for different study participants. Fig. 10a shows the place attachment surface for participant 362199345, which was detailed in Fig. 9. In Fig. 10b, participant 832284465 described seven fishing areas and the “waterfall” that cascades from the lower dam spillway (furthest to the east). Some of the feature surfaces described by this participant merge together because of their proximity. Fig. 10c comes from participant 796769349, who focused on the trails and forested areas of the park but paid less attention to aquatic features. Fig. 10d (participant 046110296) focuses on a natural chute that carries water between the lakes, the “Big Rock” and the footbridge to the west of the upper lake. These were marked on the optional sketch map provided by the participant. The summed fuzzy surface for all participants yields some unexpected results (Fig. 11). Although the entire park area has a place attachment that is medium or higher, we see several significant areas outside the park boundary that are important. The first of these is a forested area with medium values to the northwest of the park. To the southwest of the park, across the Nanaimo Parkway, another large forested region has medium and high values. In the southeast corner of the study area, an area of undeveloped private land has medium values. Finally, to the south, a large, city-owned lot contains areas of medium and high place attachment. We can see from this that “Colliery Dam Park” as viewed by the study participants extends beyond its official boundaries. Most study participants were unaware that the city-owned lot to the south was not legally part of the park. The Nanaimo Parkway Trail runs through this lot, and there is no delineation of the park boundary in the area. Recently, Nanaimo City Council voted to add this lot to the park (Cunningham, 2017). Another finding is the relative importance of the upper (western) lake, relative to the lower (eastern) lake. The upper lake is nearly encircled by an area of very high place attachment whereas the lower lake has a few areas of high place attachment along its southern and eastern
Aij is the value of the place attachment surface; A is the place attachment value for the feature; Amin is the minimum possible place attachment value; and Mij is the multiplier value of the composite surface from Equation (3). Place attachment surfaces for different participants are difficult to examine together, so we normalize the results, so that they may be compared or combined. Given a theoretical place attachment interval of [1.7 …, 13.32 ] from Equation (1), we can normalize each participant's place attachment surface to create a fuzzy surface with the values [0.0, 1.0] (Equation (5)).
Fij = ( Aij − min (Aij ) ) / ( max (Aij ) − min (Aij ) )
(5)
Where: Fij is the value [0.0, 1.0] for each pixel in the fuzzy surface raster Aij is the value [1.7 …, 13.32 ] for each pixel in the place attachment raster The maximum value of each fuzzy surface will now reach 1.0 only if a participant reports the highest possible place attachment value (13.32 ). 5. Results Using the methods that we have described, we are able to create visualizations of people's sense of place attachment. We created place attachment surfaces for each of the 239 valid participants, which helped us to understand the unique perspectives of each individual. Vague concepts can now be mapped and made visible as place attachment surfaces. Place attachment surfaces can be complex and difficult to describe except in map form. Massey (1997, p. 321) states, “If it is now recognized that people have multiple identities then the same point can
Fig. 11. Summed fuzzy surfaces for all study participants. 85
Applied Geography 99 (2018) 77–88
B. Maguire, B. Klinkenberg
memories will be recalled. While it is possible for an individual to recall memories of a place anywhere, the prevalence of reminders will tend to decrease with distance. 7. A theoretical model of place attachment decay is sufficient for modeling.
shores. The low place attachment in this area may be because the children and youth who frequent this area were under-represented in the sampling. 6. Discussion
6.3. Limitations of approach
The results demonstrate how the Place Analysis System can be used to collect, store, process and visualize participants’ attachment to place. The approach that we have described has many strengths and some weaknesses.
There are a few limitations to the approach that we employed in this study. Identifying these limitations and developing solutions will help improve the effectiveness of our approach in future. Constructing the initial catalog of mapped features in the park was very time consuming, even though some data were available. We mapped many features that were never mentioned and missed a few that were described by participants. Despite its advantages, in-situ data collection is logistically challenging. Brown and Weber (2011, p. 2) point out that park visitors “ … are intent on enjoying limited leisure time. Further, parks are often large and dispersed … which makes intercepting park visitors difficult.” Practical considerations limited our data collection to six well-traveled sites within the park. Unlike the other processing and analysis steps used in the PAS, in-situ data collection does not scale well with increases to the size of the study area or the population sampled. At present, the PAS provides a first approximation of place attachment levels throughout the study area. Place attachment surfaces are calculated based on the distance from nearby features; where there are numerous features, the surface will have greater detail than in places where there are few. However, in most cases the cells in the raster will be calculated based on Equation (4); only if a cell is occupied will its value come directly from a feature. Since it is impractical to ask participants about place attachment for each cell in the study area - every 2 m (6.6 ft) - having a calculated surface is a necessity. Several anonymous reviewers commented on the disconnect between the measures used in this study and the concepts of place dependence and place identity used in the literature. While the measures used provide a rough measure of the concepts, better measures have been demonstrated in the literature and will be employed in future studies. Although the place attachment surfaces for individuals and groups can help people to understand points of agreement and disagreement in their understanding of place, people are accustomed to dealing with solid, unambiguous lines on maps. To avoid confusion, we need some way to establish linear boundaries for the place attachment surfaces. Although qualitative data in the form of comments, descriptions, sketch maps and participant photographs have been collected in the PAS, they are not well integrated with the analysis and display of results. The ability to consult qualitative data easily while performing analysis can allow for a more nuanced understanding of the data. At present, the PAS is still under development. The PAS has been used in a park environment, but not in other environments where place attachment is important. As the system is refined, it will become sufficiently stable to release to other researchers. With any such release, user documentation and a training program will need to be developed.
6.1. General approach Although collecting data in-situ is difficult, some of the biases of other sampling methods are avoided. For example, some sampling methods, such as mail-in or internet surveys will tend to over-represent established long-term residents and lists of potential participants drawn from property registers will miss renters, the homeless and visitors. A random selection of park users reduced the likelihood of these biases occurring, however there is no guarantee that any sampling procedure is fully representative. The method ties place attachment surfaces to accurately mapped features so that they can be compared with each other and analyzed as a group. This permits comparative analysis of the place attachment surfaces in a GIS. We can compare differences in the strength and extent of place attachment for individual features and for the place as a whole. Using this system, the values at any point in a place attachment surface can be explained, given the features chosen, emotional intensities, importance values, and awareness distances provided by the participants. 6.2. Assumptions Because place attachment is a complex, multidimensional concept which is not yet well defined (Jorgensen & Stedman, 2001), producing a working model of the concept is challenging. As one anonymous reviewer pointed out, balancing the desire to model place attachment perfectly and the need to collect data from the general population in an understandable manner creates an inherent tension in the software design. Our approach has been to model place attachment from the bottom-up, developing a modest model that has sufficient predictive power to be useful. In doing this, we have had to make several assumptions in our research methodology. 1. Our research is reductionist in approach and assumes that it is possible to model a place as the sum of the individual features that make the place important. We consider genius loci to be an emergent property of an individual's interactions with the features that make up a place. 2. The model assumes that place dependence and place identity contribute equally to the concept of place attachment. Given that these are separate psychological constructs (Williams, 2016), we give them equal value by default. 3. Given that multiple questions have high overlap with place dependence and others have a high overlap with place identity (Vaske & Kobrin, 2001; Williams & Vaske, 2003), a single well-chosen question can effectively represent each of the concepts. 4. The intensity of an emotion provided by a participant is a reasonable proxy for place identity, and the importance of place is a reasonable proxy for place dependence. 5. Place attachment is a continuous phenomenon that extends from discrete features that are accurately mapped. Since people retain memories of places and these become part of a person's self-identity (Pred, 1984), there is no discrete boundary at which place attachment abruptly ceases. 6. Distance is the dominant factor controlling the likelihood that
6.4. Future steps Improvements to our methods follow from the limitations listed above and can be broken into procedural and software changes. The procedural changes include changes to the survey methodology and the data collected. In future, we will adopt a rapid appraisal methodology (Gunderson & Watson, 2007), including key informant interviews and snowball sampling (Lowery & Morse, 2013). This will allow us to use the knowledge of long-term residents to build an appropriately sized inventory of mapped features before surveying commences. Such an inventory should more closely match the features that participants 86
Applied Geography 99 (2018) 77–88
B. Maguire, B. Klinkenberg
As the software is developed, it will be made available to other researchers. To support this initiative, a training program will be developed to teach researchers how to configure the system, collect data for the PAS, enter the data, and produce models of place attachment. We would like to support a group of researchers as they systematically examine the dynamics of some of the interesting places around the world.
mention, making the surveying process more efficient. A rapid appraisal methodology will also help to assess the general level of interest about the place being studied and will provide advance notice if survey participation needs to be incentivized. Another way to improve the response rate is by making the data collection procedures easier and better aligned with the concepts of place identity and place dependence. We propose adopting a single question for place identity and place dependence from those listed in Vaske and Kobrin (2001) and Williams and Vaske (2003). Using a more direct method to measure place identity and place dependence would avoid the issue of emotions being culturally defined constructs (Russell, 2003) and the level of English fluency and time required to choose a single emotion from a lengthy list. To mitigate some of the disadvantages of in-situ sampling, future paper questionnaires will be enhanced and modified so that they can be filled out on a clipboard as participants travel around a place. The survey form will also be made more accessible by increasing the font size for those who have difficulty reading and by making the layout more appropriate for use in oral interviews. These measures will improve the amount and quality of data returned. Paper questionnaires can also be supplemented with a smartphone application that could significantly improve the spatial, temporal and qualitative aspects of data collection. A well-designed smartphone application can improve the quality of data collected, make the process easier and more enjoyable, and automate aspects of the process, so that data collection is accessible to more people. Because a GPS-based application can record multiple points, these might help to define place attachment surfaces with greater detail. Another benefit is that data collection efforts can be ongoing over days or months. Such a technology can be used in conjunction with paper-based data collection techniques to maintain our ability to reach a broad cross section of park users. Because the PAS makes use of points, lines, and polygons, it can be used to map virtually any discrete feature (Heywood, Cornelius, & Carver, 2011), from which a continuous place attachment surface can be constructed. Furthermore, using all three types of features avoids limiting the analysis of places to any particular scale. For example, a distinctive building that serves as a landmark might be represented as a polygon in a small study area or as a point in a medium or large study area. Some features that are relatively important in small study areas might not be so important in larger study areas, so might not be shown at all. Being able to use points, lines, and polygons provides flexibility to show features appropriately at different scales. Applying our methods to other types of places will help us to identify methodological limitations and improve the design of the software. It would be interesting to examine people's place attachment to places other than parks. As the methodology and software are developed, we might examine cities, communities or neighborhoods, important religious and secular places, historical sites, or even “nonplaces” such as airports, malls, and suburbs (Augé, 1992). In addition to the procedural changes, there are several changes to the software that can be implemented. In future studies, some changes will be made to the place attachment surfaces. The current isotropic place attachment surfaces can be improved by modeling the effect of physical barriers (e.g. roads, fences, hills) that limit travel into certain areas. It would also be valuable to define a way to defuzzify the boundaries of a place attachment surface. An algorithm needs to be developed to create “socially acceptable” discrete boundaries that are compatible with other GIS models. We plan to further integrate qualitative analysis into the PAS. This will help to triangulate the results of quantitative analysis, augmenting the existing support for the model. In addition, the quantitative and qualitative information will be tied to the map of the study area in the GIS. This will help to further improve the contextualization of the results produced by the PAS, allowing us to examine the source data for the place attachment surfaces with a mouse click.
6.5. Implications Being able to determine the spatial distribution of place attachment yields new insights on the importance of place, not only for the entire study area, but also for subsections. For our current study area, we can combine place attachment surfaces to determine the collective place attachment for distinct classes of park users (runners, cyclists, fishers, swimmers), different demographic classes (age, ethnicity), and distinct groups by time of visitation. These types of analyses can provide new insights about how places work. Much theoretical work has been done, and representing the spatial distribution of place attachment is a logical extension of these methods. Mapping place attachment makes it more concrete, aiding place research and promoting better-informed decisions. Much remains to be done now that we have three-dimensional surfaces that show place attachment. Place attachment surfaces can serve as a foundation for future academic studies of place, and the methods and technology should be easily adaptable for use in disciplines such as planning, where place attachment plays a significant role. Acknowledgments Stuart Dixon and Chris Mueller provided invaluable help with data collection for this project. The City of Nanaimo is recognized for its open data access policy. Special thanks also go out to our advisors and reviewers who provided much advice and feedback. Funding: Vancouver Island University helped fund this project through its student work opportunity program. The university was not involved in the study design, data collection, interpretation, or analysis of data. References Augé, M. (1992). Non-Places: Introduction to and anthropology of supermodernity. London: Verso. Brown, G. (2005). Mapping spatial attributes in survey research for natural resource management: Methods and applications. Society & Natural Resources, 18, 17–39. Brown, G., & Raymond, C. (2007). The relationship between place attachment and landscape values: Toward mapping place attachment. Applied Geography, 27(2), 89–111. Brown, G., Raymond, C., & Corcoran, J. (2014). Mapping and measuring place attachment. Applied Geography, 57, 42–53. Brown, G., & Weber, D. (2011). Public participation GIS: A new method for national park planning. Landscape and Urban Planning, 102, 1–15. Burrough, P. (1996). Natural objects with indeterminate boundaries. In P. Burrough, & A. Frank (Eds.). Geographic objects with indeterminate boundaries (pp. 3–28). London: Taylor and Francis. Carver, S., Watson, A., Waters, T., Matt, R., Gunderson, K., & Davis, B. (2009). Developing computer-based participatory approaches to mapping landscape values for landscape and resource management. In S. Geertman, & J. Stillwell (Eds.). Planning support systems best practice and new methods (pp. 431–448). New York: Springer Science +Business Media. Cunningham, T. (2017). City council adds park land at Colliery dams. Nanaimo News Bulletin, 29(23), 1. Jul. 27, 2017 http://www.nanaimobulletin.com/news/citycouncil-adds-park-land-at-colliery-dams/. Dornič, S. (1967). Reports from the psychological laboratoriesSubjective distance and emotional involvement: A verification of the exponent invarianceVol. 237. Stockholm: University of Stockholm. Ekman, G., & Bratfisch, O. (1965). Subjective distance and emotional involvement: A psychological mechanism. Acta Psychologica, 24, 430–437. Evans, A., & Waters, T. (2007). Mapping vernacular geography: web-based GIS tools for capturing "fuzzy" or "vague" entities. International Journal of Technology, Policy and Management, 7(2), 134–150. Fonte, C. (2008). Geographical surfaces as entities. In W. Lodwick (Ed.). Fuzzy surfaces in GIS: Theory, analytical methods, algorithms, and applications (pp. 63–84). Boca Raton,
87
Applied Geography 99 (2018) 77–88
B. Maguire, B. Klinkenberg
Arnold. Plutchik, R. (1962). The emotions: Facts, theories, and a new model. New York: Random House. Plutchik, R. (1980). Emotion: A psychoevolutionary synthesis. New York: Harper and Row. Pred, A. (1984). Place as historically contingent Process: Structuration and the timegeography of becoming places. Annals of the AAG, 74(2), 279–297. Russell, J. (2003). Core affect and the psychological construction of emotion. Psychological Review, 110(1), 145–172. Stedman, R. (2002). Toward a social psychology of place: Predicting behavior from placebased cognitions, attitude, and identity. Environment and Behavior, 34(5), 561–581. Thrift, N. (1994). Inhuman Geographies: Landscapes of speed, light and power. In P. Cloke, M. Doel, D. Matless, M. Phillips, & N. Thrift (Eds.). Writing in the rural: Five cultural geographies (pp. 191–248). London: Paul Chapman. Vaske, J., & Kobrin, K. (2001). Place attachment and environmentally responsible behavior. The Journal of Environmental Education, 32(4), 16–21. Williams, D. (2014). Making sense of 'place': Reflections on pluralism and positionality in place research. Landscape and Urban Planning, 131, 74–82. Williams, D. (2016). Personal communication. June 27, 2016. Williams, D., & Vaske, J. (2003). The measurement of place attachment: Validity and generalisability of a psychometric approach. Forest Science, 49(6), 830–840. Zadeh, L. (1965). Fuzzy sets. Information and Control, 8, 338–353.
FL: CRC Press. Fonte, C., & Lodwick, W. (2005). Modelling the fuzzy spatial extent of geographical entities. In F. Petry, V. Robinson, & M. Cobb (Eds.). Fuzzy modelling with spatial information for geographical problems (pp. 121–142). Berlin: Springer. Golledge, R. (1978). Representing, interpreting and using cognized environments. Papers of the regional science association. Gould, P., & White, R. (1992). Mental maps (2nd ed.). New York: Routledge. Gunderson, K., & Watson, A. (2007). Understanding place meanings in the bitterroot national forest, Montana. Society & Natural Resources, 20, 705–721. Heywood, I., Cornelius, S., & Carver, S. (2011). An introduction to geographical information systems (4th ed.). Harlow, Essex: Pearson. Jorgensen, B., & Stedman, R. (2001). Sense of place as an attitude: Lakeshore owners attitudes toward their properties. Journal of Environmental Psychology, 21, 233–248. Kitchin, R., & Blades, M. (2002). The cognition of geographic space. London: I.B. Tauris. Lowery, D., & Morse, W. (2013). A qualitative method for collecting spatial data on important places for recreation, livelihoods, and ecological meanings: Integrating focus groups with public participation geographic information systems. Society & Natural Resources, 26, 1422–1437. Lynch, K. (1960). The image of the city. Cambridge, MA: MIT Press. Massey, D. (1997). A global sense of place. In T. Barnes, & D. Gregory (Eds.). Reading human geography: The poetics and politics of inquiry (pp. 315–323). London: Edwin
88
Contents lists available at ScienceDirect
Applied Geography journal homepage: www.elsevier.com/locate/apgeog
Visualization of place attachment a,∗
Brad Maguire , Brian Klinkenberg a b
T
b
Department of Geography, Vancouver Island University, 900 Fifth Street, Nanaimo, BC, V9R 5S5, Canada Department of Geography, University of British Columbia, 1984 West Mall, Vancouver, BC, V6T 1Z2, Canada
A R T I C LE I N FO
A B S T R A C T
Keywords: GIS Mapping Place Place attachment Sense of place Visualization
The concept of place attachment is amorphous and difficult to study. To address this, we set out to map and analyze place attachment for a city park in Nanaimo, British Columbia, Canada. We developed a GIS application called the Place Analysis System (PAS) to allow place attachment data to be collected, georegistered, stored and processed into map form. Upon analysis, we found notable differences in the spatial distribution of place attachment between groups and under different weather conditions. The ability to visualize and analyze place attachment can assist with both academic and applied studies in fields such as planning.
1. Introduction Recent research into place attachment has revealed the major components of this concept. Nevertheless, place attachment remains ethereal, lacking in form and context. Despite extensive study in the geographic literature, little attention has been given to the spatial distribution of place attachment. We addressed this problem by connecting place attachment to realworld places. Using a geographic information system (GIS), we mapped the strength and spatial distribution of place attachment around the features that make up places. This makes place attachment easier to visualize and measure, making it more amenable to empirical analysis and applied research. 1.1. Understanding place Place studies can be complicated because there are many concepts of place from different fields of study and intellectual traditions. Williams (2014) defines three views of place: demos, which views place as being dynamic, cosmopolitan, and spatiotemporal, ethnos, in which place is defined by group identities and parochial attachments to place, and bios, in which place is defined by bioregionalism and a “back to the land” ethic. Our research examines place primarily from the demos perspective. Many of the concepts used to study place benefit from the work of Massey (1997), Pred (1984) and Thrift (1994). This perspective re-
∗
cognizes the cosmopolitan nature of place and leads us to model place as unbounded and connected to other places. The dynamic nature of place can be analyzed by studying where people travel at different times and under different weather conditions. 1.2. Modeling place using GIS In this paper, we introduce and demonstrate the use of the Place Analysis System (PAS), a GIS application that facilitates improved data collection and analysis of place based on place attachment theory (Stedman, 2002; Williams & Vaske, 2003). The PAS enables researchers to create fuzzy models of place attachment, which facilitate the understanding, analysis and communication of this concept. We will demonstrate how the system can help visualize place attachment, taking study participants’ vague notions of place attachment and making them concrete. 1.3. Study area We chose Colliery Dam Park, a 27.2 ha (67.2 ac.) semi-wilderness park in Nanaimo, Canada for our study area (Fig. 1). The centerpieces of the park are two artificial lakes formed by the Colliery Dams (Fig. 2). These lakes formerly supplied water for coal mining, but now serve recreational purposes (Fig. 3). The park has much in common with larger parks such as Central Park in New York, Rock Creek Park in Washington, D.C. or Stanley Park in Vancouver.
Corresponding author. E-mail address: [email protected] (B. Maguire).
https://doi.org/10.1016/j.apgeog.2018.07.007 Received 30 August 2017; Received in revised form 28 June 2018; Accepted 2 July 2018 0143-6228/ © 2018 Elsevier Ltd. All rights reserved.
Applied Geography 99 (2018) 77–88
B. Maguire, B. Klinkenberg
now moving towards decentralized or mobile computer data entry through web pages (Carver et al., 2009), and smartphone applications. In these GIS studies, places have been conceptualized in different ways. They have been represented as points (Brown & Raymond, 2007; Brown, 2005; Brown et al., 2014), polygons (Lowery & Morse, 2013), or rasters. Carver et al. (2009) developed a method that uses a “spray paint” metaphor, in which a participant can “paint” place importance onto a raster map. This technique allows for the direct recording of a fuzzy surface.
Abbreviations PAS
Place Attachment System, the GIS application used to collect and organize data for this study
2. Literature review 2.1. Working with fuzzy surfaces
2.4. Research advances
Individuals view places differently from one another (Williams, 2014) and have different features and levels of interest. We can represent these as fuzzy surfaces (Burrough, 1996; Fonte & Lodwick, 2005; Fonte, 2008; Zadeh, 1965), in which place attachment decays as distance increases. This decay represents the decrease in awareness as features fade from short-term memory. The awareness never reaches zero because long-term memories of a feature may be recalled at any time with the right stimuli.
The method proposed in this paper offers a general solution for the modeling of place attachment based on the demos concept of place. Collecting the data soon after participants have visited the features allows us to work with participants while impressions of new features are fresh. In addition, the sights, sounds, and smells of the park assist in the recollection of past memories of park features. As well, participants can use body language to indicate the distance, direction, size and configuration of the features that they are describing. Brown (2005) mentioned that “… in the Chugach National Forest study, many of the intrinsic value dots were placed on empty spaces …” (p. 33). Although Brown attributes this to people recognizing the intrinsic value of wild places, another possibility is that a lack of familiarity with certain places or a lack of map literacy caused people to place the dots inaccurately. Brown also relates that the points were used to infer polygon boundaries based on the density of their placement, which is a subjective process. Written and verbal communications were used when interacting with the participants to avoid issues related to computer literacy. Because we mapped the features in advance, we were able to use placebased nomenclature (Evans & Waters, 2007; cited in Carver et al., 2009) to match the names of features with their mapped counterparts. This allowed features to be identified without having to rely on the cartographic or map reading skills of the participants. Once identified, the features were automatically copied for analysis so that the original data remained intact. The decay in place attachment was calculated for each feature, and this was used to create feature surfaces that describe how memories of features decrease as distance increases. Rather than using a pre-defined interpolation method (e.g. spline, inverse distance weighting), the shape of each feature surface was based on the decay rate of place memories, as described by Dornič
2.2. Modeling emotion We employed Plutchik's (1962, 1980) psychoevolutionary model of the emotions as the dictionary of emotions used in this study. A numerical intensity value is associated with each emotion in Plutchik's model. This value is used in the PAS; the emotion name simply provides a convenient reference for the participants to use. 2.3. Efforts to map place Geographers and psychologists have been refining the concept of place over the past six decades. Work on “mental maps” has revealed the distortions present in how people model their surroundings (Gould & White, 1992). These distortions make it difficult to operationalize findings, because it is difficult to identify exactly where features are located (Golledge, 1978, p. 41). The work of Lynch (1960) has been built upon and expanded by modern researchers using GIS. There has been a gradual shift from paper-based to computer-based methods of recording spatial data. Early work, such as Brown (2005) made use of paper-based methods for data collection, whereas later work (Brown & Raymond, 2007; Brown, Raymond, & Corcoran, 2014) made use of computer data entry. We are
Fig. 1. The study area around Colliery Dam Park.
78
Applied Geography 99 (2018) 77–88
B. Maguire, B. Klinkenberg
Fig. 2. Colliery Dam Park and the study area.
Fig. 3. The lower lake and dam at Colliery Dam Park (photo by B. Maguire).
submissions were skipped. Out of 302 responses, 63 were rejected, leaving 239 responses for analysis (Table 1). A Kruskal-Wallis H test was used to evaluate whether the three different sampling methods yielded similar place attachment values. An examination of boxplots showed similar distributions in the place attachment values, which was confirmed by test results that showed similar overall place attachment values (H(2) = 3.091 p = .213). Based on these findings, we are confident that all three sampling methods can be combined to create an overall composite view of place attachment for Colliery Dam Park.
(1967). We term this a theory-based interpolation method. The software was designed to combine feature surfaces for related objects. For example, if a participant describes “water features” we can combine all the lakes (polygons) and rivers (lines) to create a single composite feature. From this, we construct a composite surface, which is simply the combination of all component feature surfaces. In turn, the composite surfaces for all features are combined to create a single place attachment surface for each participant. 3. Data and methods 3.1. Sampling strategy
Table 1 Numbers of participants by sampling method and rejection status.
We collected data from six sampling points at well-traveled locations from Oct. 20, 2011 to Sept. 29, 2012 (Fig. 2). The day of the week (including weekends) and time of collection (morning, afternoon, or evening) were rotated to maximize the chances of interviewing people from different sub-populations of users. Samples were mostly collected randomly, except at times of low traffic, when convenience sampling was employed. A considerable number of park users volunteered to take the survey. In cases where participants provided invalid information or demonstrated that they did not understand the questions, their
Not Rejected Rejected Total
79
Randomly Sampled
Volunteered
Convenience Sampled
Total
142 (47.0%) 32 (10.6%) 174 (57.6%)
85 (28.1%) 26 (8.6%) 111 (36.8%)
12 (4.0%) 5 (1.7%) 17 (5.6%)
239 (79.1%) 63 (20.9%) 302 (100.0%)
Applied Geography 99 (2018) 77–88
B. Maguire, B. Klinkenberg
Fig. 4. Main survey questionnaire as filled out by participant 362199345.
To help participants complete the questionnaire, they were provided with a double-sided reference card on which Awareness Distance was explained (Fig. 5). This card also listed 53 emotions in addition to “no emotion.” Thirty-two of these were shown on Plutchik's wheel of emotions (Plutchik, 1962), and those remaining were shown in a table. If a participant mentioned an emotion that was not described on the reference card, the software allowed any of the 118 emotions described by Plutchik (1962) to be chosen. For each feature, participants were invited to provide qualitative information including descriptions of the feature, photographs, sketch maps, or explanations of why it was important. These sources of information provided context for the place attachment surfaces produced by the PAS. Participants were also asked to provide their age, gender, ethnicity, occupation, level of education, and postal code. This demographic information showed that study participants were divided roughly equally between genders and age groups, except for minors (under age 19) who required parental assent before they could participate (Table 2). Based on the self-reporting of educational status, 55 participants (18.2%) had an elementary education, 39 (12.9%) had a secondary education, 165 (54.6%) were university or college graduates, and 43 (14.2%) had post-secondary degrees. On the final page, the survey taker noted the date, time, weather conditions and location of data collection. The time of collection is associated with all data, so that spatiotemporal changes to place attachment can be examined.
3.2. Data collection Participants were asked to fill out a paper questionnaire to identify an open-ended number of park features (Fig. 4). After providing the feature name, they were asked to indicate the main emotion associated with the feature, which was used to determine the emotional intensity. Next, participants rated the importance of the feature on a 7-point Likert scale (1 = very unimportant, 7 = very important). The importance was used as an indicator of place dependence and the intensity of the emotion was used as an indicator of place identity. Participants next indicated the distance at which they were no longer aware of the feature. The awareness distance is defined as “How close you must be before you become aware of it [a feature] when approaching, or how far away you must be before you lose awareness of it when you are leaving.” Awareness distance (the XY value) measures the Euclidean distance from the nearest component of a composite feature and is used to calculate the decay of place attachment (the Z value) as distance increases. It extends beyond the visible range to where the feature is no longer in short-term memory. For first-time visitors to a place, awareness distance is the same as visual distance, but with repeated visits, place attachment increases, knowledge of the place grows, and a person's mental map becomes more complete (Gould & White, 1992; Kitchin & Blades, 2002). Through this process, people gain configurational knowledge of important features that are not visible from particular locations. The awareness distance is uncorrelated with the size of the feature. A single point feature can have a larger awareness distance than a large polygon feature. It should be noted that in such a case, the total area within the awareness distance of the polygon feature may be greater than that within the awareness distance of the point, simply because the polygon has a larger area than the point.
3.3. Data entry A researcher entered data from the paper form into the PAS. The PAS interface (Fig. 6) mimics the paper form to facilitate data entry and reduce transcription errors. Validity checks are performed (e.g., 80
Applied Geography 99 (2018) 77–88
B. Maguire, B. Klinkenberg
Fig. 5. The reference card that was provided to participants to assist them with filling in the survey.
Table 2 Breakdown of study participants by age and gender. Age
Unreported
< 19
19–29
30–39
40–49
50–59
> 60
Total
Males
22 (7.3%) 9 (3.0%) 31 (10.3%)
4 (1.3%) 5 (1.7%) 9 (3.0%)
33 (10.9%) 29 (9.6%) 62 (20.5%)
21 (7.0%) 18 (6.0%) 39 (12.9%)
19 (6.3%) 26 (8.6%) 45 (14.9%)
24 (7.9%) 36 (11.9%) 60 (19.9%)
31 (10.3%) 25 (8.3%) 56 (18.5%)
154 (51.0%) 148 (49.0%) 302 (100.0%)
Females Total
method used. All three tables use a shared 9-digit random key to allow the tables to be joined for analysis, while preserving the anonymity of study participants.
incorrect postal codes are rejected) to help ensure that data are entered correctly. To enable the use of place-based nomenclature, each named feature is matched with one or more previously mapped features. The software presents a list of 1322 mapped features to choose from as well as a series of buttons to help the researcher choose composite features (Fig. 7).
3.5. Data processing Once the data have been stored in the database, they can be processed to build a place attachment surface for each participant. For each feature, a feature surface is created as a floating-point raster with a 2.5 m (8.2 ft) resolution, which maintains detail, but still allows for efficient data processing. Although the centering of the point, line, and polygon features is exact, the curvature of the feature surfaces is theoretical, based on the work of Dornič (1967) (Equation (2)). The feature surfaces are then combined using a Fuzzy OR function (Zadeh, 1965) to create a composite surface for each composite feature (Fig. 8). When feature surfaces overlap, the composite surface maintains the highest input value.
3.4. Organization of data Once the data have been entered and validated, the system automatically transfers it to a relational database for analysis. Three main tables are used in the database. Mainform_Features stores information about the features described. Mainform_Participant stores information about the participants, such as frequency of park visits and demographic data. Mainform_Questionnaire stores the time, date and location of data collection, the weather conditions, and the sampling 81
Applied Geography 99 (2018) 77–88
B. Maguire, B. Klinkenberg
Fig. 6. An analog of the paper form reduces the likelihood of transcription errors.
All composite features are then combined using a Fuzzy OR function to create a place attachment surface for the participant. Fig. 9 provides an example of how a place attachment surface is built. A planimetric view of this place attachment surface is shown in Fig. 10a. The place attachment surface is scaled to a range of [0.0, 1.0] so
We next scale the Z-value of the composite surface [0.0, 1.0] by the place attachment value for each feature (Equation (4)). The scaled composite surface ranges from the place attachment value at the feature outward to the minimum place attachment value (1.7 …) at and beyond the awareness distance.
Fig. 7. The Choose Feature window allows mapped features to be assigned to each feature listed by a participant. The buttons help to quickly select individual or composite features. 82
Applied Geography 99 (2018) 77–88
B. Maguire, B. Klinkenberg
Fig. 8. Composite surfaces for point, line and polygon features, which are shown in black on top of each surface. The lower line for each surface shows the awareness distance. A composite feature may be composed of multiple elements, such as the boardwalk in the center image. Features may also have holes, such as those in the polygon on the right.
Where:
that it can be compared and combined with other participants’ results (Equation (5)). The range [1.7 …,11.74] in Fig. 9 is scaled to [0.00, 0.86] because the place attachment surface has the lowest possible minimum value (1.7 …), but not the highest possible maximum value (13.32) . Because each fuzzy surface extends away from features that have been accurately mapped, the surfaces for different participants can be measured, compared and summed together. This allows a summed fuzzy surface to show the “consensus view” of a group or the entire population of study participants.
A is the place attachment value; I is the place identity; and D is the place dependence of the feature. Substituting the minimum and maximum place identity and place dependence values into Equation (1) produces an interval of [1.7 …, 13.32 ] for place attachment values. The curve of the decay surface is based on research about how memories of emotions decay with distance. Ekman and Bratfisch (1965) found that the emotions related to a place follow an inverse square law, but the exponent was later refined by Dornič (1967) to yield Equation (2):
4. Theory/calculation Plutchik's intensity values are based on empirical research, in which a group of research subjects was asked to rate the psychological impact of many emotions, from 1 (low) to 11 (high). Plutchik (1980) found that emotions with intensity values below 3 were indistinguishable from one another, so values below 3 were not used. Combining the [3,11] interval for place identity (intensity) with the [1,7] interval for place dependence, we come up with the following formula for place attachment, in which place identity (9 values) is separate, but equal in importance to place dependence (7 values):
A = 7/9 (I − 2) + D
Zij = dij−0.47
(2)
Where: Zij represents the value of the decay surface; and dij is the distance from the feature. Because the function is asymptotic to zero, it supports Massey's (1997) assertion that places have no boundaries since such a surface extends around the globe. Equation (3) creates a normalized
(1)
Fig. 9. Place attachment surface for participant 362199345 viewed from the northeast, showing the contributions made by individual composite surfaces. The place attachment value shown for each composite surface is A, and the awareness distance is d. The minimum place attachment value has been removed for clarity.
83
Applied Geography 99 (2018) 77–88
B. Maguire, B. Klinkenberg
Fig. 10. Place attachment surfaces for four participants. Darker (redder) values represent more important areas, whereas lighter (bluer) values represent less important ones.
dawareness is the awareness distance expressed by the study participant; and dmin is the minimum possible distance.
logarithmic surface with the correct curvature, but which has a value of 1.0 at the center and 0.0 at and beyond the distance specified by the participant.
if dij = 0 ⎫ ⎧1 ⎪ ⎪ dij−0.47 − d awareness−0.47 > ≤ if d 0 and d d Mij = ij ij awareness ⎬ ⎨ dmin−0.47 − d awareness−0.47 ⎪ ⎪0 if dij > dawareness ⎭ ⎩
To create a place attachment surface, the composite surface produced in Equation (3) is scaled to ensure that values range from the feature's place attachment value (A) at the inner edge of the feature surface to the minimum possible place attachment value (Amin) at and beyond the awareness distance (Equation (4)).
(3)
Where:
Aij = (A − Amin ) Mij + Amin
Mij is the multiplier value for the place attachment surface; dij is the Euclidean distance recorded in each pixel;
Where:
84
(4)
Applied Geography 99 (2018) 77–88
B. Maguire, B. Klinkenberg
be made in relation to places.” We can see graphical evidence of this in Fig. 10, which shows a selection of four place attachment surfaces constructed for different study participants. Fig. 10a shows the place attachment surface for participant 362199345, which was detailed in Fig. 9. In Fig. 10b, participant 832284465 described seven fishing areas and the “waterfall” that cascades from the lower dam spillway (furthest to the east). Some of the feature surfaces described by this participant merge together because of their proximity. Fig. 10c comes from participant 796769349, who focused on the trails and forested areas of the park but paid less attention to aquatic features. Fig. 10d (participant 046110296) focuses on a natural chute that carries water between the lakes, the “Big Rock” and the footbridge to the west of the upper lake. These were marked on the optional sketch map provided by the participant. The summed fuzzy surface for all participants yields some unexpected results (Fig. 11). Although the entire park area has a place attachment that is medium or higher, we see several significant areas outside the park boundary that are important. The first of these is a forested area with medium values to the northwest of the park. To the southwest of the park, across the Nanaimo Parkway, another large forested region has medium and high values. In the southeast corner of the study area, an area of undeveloped private land has medium values. Finally, to the south, a large, city-owned lot contains areas of medium and high place attachment. We can see from this that “Colliery Dam Park” as viewed by the study participants extends beyond its official boundaries. Most study participants were unaware that the city-owned lot to the south was not legally part of the park. The Nanaimo Parkway Trail runs through this lot, and there is no delineation of the park boundary in the area. Recently, Nanaimo City Council voted to add this lot to the park (Cunningham, 2017). Another finding is the relative importance of the upper (western) lake, relative to the lower (eastern) lake. The upper lake is nearly encircled by an area of very high place attachment whereas the lower lake has a few areas of high place attachment along its southern and eastern
Aij is the value of the place attachment surface; A is the place attachment value for the feature; Amin is the minimum possible place attachment value; and Mij is the multiplier value of the composite surface from Equation (3). Place attachment surfaces for different participants are difficult to examine together, so we normalize the results, so that they may be compared or combined. Given a theoretical place attachment interval of [1.7 …, 13.32 ] from Equation (1), we can normalize each participant's place attachment surface to create a fuzzy surface with the values [0.0, 1.0] (Equation (5)).
Fij = ( Aij − min (Aij ) ) / ( max (Aij ) − min (Aij ) )
(5)
Where: Fij is the value [0.0, 1.0] for each pixel in the fuzzy surface raster Aij is the value [1.7 …, 13.32 ] for each pixel in the place attachment raster The maximum value of each fuzzy surface will now reach 1.0 only if a participant reports the highest possible place attachment value (13.32 ). 5. Results Using the methods that we have described, we are able to create visualizations of people's sense of place attachment. We created place attachment surfaces for each of the 239 valid participants, which helped us to understand the unique perspectives of each individual. Vague concepts can now be mapped and made visible as place attachment surfaces. Place attachment surfaces can be complex and difficult to describe except in map form. Massey (1997, p. 321) states, “If it is now recognized that people have multiple identities then the same point can
Fig. 11. Summed fuzzy surfaces for all study participants. 85
Applied Geography 99 (2018) 77–88
B. Maguire, B. Klinkenberg
memories will be recalled. While it is possible for an individual to recall memories of a place anywhere, the prevalence of reminders will tend to decrease with distance. 7. A theoretical model of place attachment decay is sufficient for modeling.
shores. The low place attachment in this area may be because the children and youth who frequent this area were under-represented in the sampling. 6. Discussion
6.3. Limitations of approach
The results demonstrate how the Place Analysis System can be used to collect, store, process and visualize participants’ attachment to place. The approach that we have described has many strengths and some weaknesses.
There are a few limitations to the approach that we employed in this study. Identifying these limitations and developing solutions will help improve the effectiveness of our approach in future. Constructing the initial catalog of mapped features in the park was very time consuming, even though some data were available. We mapped many features that were never mentioned and missed a few that were described by participants. Despite its advantages, in-situ data collection is logistically challenging. Brown and Weber (2011, p. 2) point out that park visitors “ … are intent on enjoying limited leisure time. Further, parks are often large and dispersed … which makes intercepting park visitors difficult.” Practical considerations limited our data collection to six well-traveled sites within the park. Unlike the other processing and analysis steps used in the PAS, in-situ data collection does not scale well with increases to the size of the study area or the population sampled. At present, the PAS provides a first approximation of place attachment levels throughout the study area. Place attachment surfaces are calculated based on the distance from nearby features; where there are numerous features, the surface will have greater detail than in places where there are few. However, in most cases the cells in the raster will be calculated based on Equation (4); only if a cell is occupied will its value come directly from a feature. Since it is impractical to ask participants about place attachment for each cell in the study area - every 2 m (6.6 ft) - having a calculated surface is a necessity. Several anonymous reviewers commented on the disconnect between the measures used in this study and the concepts of place dependence and place identity used in the literature. While the measures used provide a rough measure of the concepts, better measures have been demonstrated in the literature and will be employed in future studies. Although the place attachment surfaces for individuals and groups can help people to understand points of agreement and disagreement in their understanding of place, people are accustomed to dealing with solid, unambiguous lines on maps. To avoid confusion, we need some way to establish linear boundaries for the place attachment surfaces. Although qualitative data in the form of comments, descriptions, sketch maps and participant photographs have been collected in the PAS, they are not well integrated with the analysis and display of results. The ability to consult qualitative data easily while performing analysis can allow for a more nuanced understanding of the data. At present, the PAS is still under development. The PAS has been used in a park environment, but not in other environments where place attachment is important. As the system is refined, it will become sufficiently stable to release to other researchers. With any such release, user documentation and a training program will need to be developed.
6.1. General approach Although collecting data in-situ is difficult, some of the biases of other sampling methods are avoided. For example, some sampling methods, such as mail-in or internet surveys will tend to over-represent established long-term residents and lists of potential participants drawn from property registers will miss renters, the homeless and visitors. A random selection of park users reduced the likelihood of these biases occurring, however there is no guarantee that any sampling procedure is fully representative. The method ties place attachment surfaces to accurately mapped features so that they can be compared with each other and analyzed as a group. This permits comparative analysis of the place attachment surfaces in a GIS. We can compare differences in the strength and extent of place attachment for individual features and for the place as a whole. Using this system, the values at any point in a place attachment surface can be explained, given the features chosen, emotional intensities, importance values, and awareness distances provided by the participants. 6.2. Assumptions Because place attachment is a complex, multidimensional concept which is not yet well defined (Jorgensen & Stedman, 2001), producing a working model of the concept is challenging. As one anonymous reviewer pointed out, balancing the desire to model place attachment perfectly and the need to collect data from the general population in an understandable manner creates an inherent tension in the software design. Our approach has been to model place attachment from the bottom-up, developing a modest model that has sufficient predictive power to be useful. In doing this, we have had to make several assumptions in our research methodology. 1. Our research is reductionist in approach and assumes that it is possible to model a place as the sum of the individual features that make the place important. We consider genius loci to be an emergent property of an individual's interactions with the features that make up a place. 2. The model assumes that place dependence and place identity contribute equally to the concept of place attachment. Given that these are separate psychological constructs (Williams, 2016), we give them equal value by default. 3. Given that multiple questions have high overlap with place dependence and others have a high overlap with place identity (Vaske & Kobrin, 2001; Williams & Vaske, 2003), a single well-chosen question can effectively represent each of the concepts. 4. The intensity of an emotion provided by a participant is a reasonable proxy for place identity, and the importance of place is a reasonable proxy for place dependence. 5. Place attachment is a continuous phenomenon that extends from discrete features that are accurately mapped. Since people retain memories of places and these become part of a person's self-identity (Pred, 1984), there is no discrete boundary at which place attachment abruptly ceases. 6. Distance is the dominant factor controlling the likelihood that
6.4. Future steps Improvements to our methods follow from the limitations listed above and can be broken into procedural and software changes. The procedural changes include changes to the survey methodology and the data collected. In future, we will adopt a rapid appraisal methodology (Gunderson & Watson, 2007), including key informant interviews and snowball sampling (Lowery & Morse, 2013). This will allow us to use the knowledge of long-term residents to build an appropriately sized inventory of mapped features before surveying commences. Such an inventory should more closely match the features that participants 86
Applied Geography 99 (2018) 77–88
B. Maguire, B. Klinkenberg
As the software is developed, it will be made available to other researchers. To support this initiative, a training program will be developed to teach researchers how to configure the system, collect data for the PAS, enter the data, and produce models of place attachment. We would like to support a group of researchers as they systematically examine the dynamics of some of the interesting places around the world.
mention, making the surveying process more efficient. A rapid appraisal methodology will also help to assess the general level of interest about the place being studied and will provide advance notice if survey participation needs to be incentivized. Another way to improve the response rate is by making the data collection procedures easier and better aligned with the concepts of place identity and place dependence. We propose adopting a single question for place identity and place dependence from those listed in Vaske and Kobrin (2001) and Williams and Vaske (2003). Using a more direct method to measure place identity and place dependence would avoid the issue of emotions being culturally defined constructs (Russell, 2003) and the level of English fluency and time required to choose a single emotion from a lengthy list. To mitigate some of the disadvantages of in-situ sampling, future paper questionnaires will be enhanced and modified so that they can be filled out on a clipboard as participants travel around a place. The survey form will also be made more accessible by increasing the font size for those who have difficulty reading and by making the layout more appropriate for use in oral interviews. These measures will improve the amount and quality of data returned. Paper questionnaires can also be supplemented with a smartphone application that could significantly improve the spatial, temporal and qualitative aspects of data collection. A well-designed smartphone application can improve the quality of data collected, make the process easier and more enjoyable, and automate aspects of the process, so that data collection is accessible to more people. Because a GPS-based application can record multiple points, these might help to define place attachment surfaces with greater detail. Another benefit is that data collection efforts can be ongoing over days or months. Such a technology can be used in conjunction with paper-based data collection techniques to maintain our ability to reach a broad cross section of park users. Because the PAS makes use of points, lines, and polygons, it can be used to map virtually any discrete feature (Heywood, Cornelius, & Carver, 2011), from which a continuous place attachment surface can be constructed. Furthermore, using all three types of features avoids limiting the analysis of places to any particular scale. For example, a distinctive building that serves as a landmark might be represented as a polygon in a small study area or as a point in a medium or large study area. Some features that are relatively important in small study areas might not be so important in larger study areas, so might not be shown at all. Being able to use points, lines, and polygons provides flexibility to show features appropriately at different scales. Applying our methods to other types of places will help us to identify methodological limitations and improve the design of the software. It would be interesting to examine people's place attachment to places other than parks. As the methodology and software are developed, we might examine cities, communities or neighborhoods, important religious and secular places, historical sites, or even “nonplaces” such as airports, malls, and suburbs (Augé, 1992). In addition to the procedural changes, there are several changes to the software that can be implemented. In future studies, some changes will be made to the place attachment surfaces. The current isotropic place attachment surfaces can be improved by modeling the effect of physical barriers (e.g. roads, fences, hills) that limit travel into certain areas. It would also be valuable to define a way to defuzzify the boundaries of a place attachment surface. An algorithm needs to be developed to create “socially acceptable” discrete boundaries that are compatible with other GIS models. We plan to further integrate qualitative analysis into the PAS. This will help to triangulate the results of quantitative analysis, augmenting the existing support for the model. In addition, the quantitative and qualitative information will be tied to the map of the study area in the GIS. This will help to further improve the contextualization of the results produced by the PAS, allowing us to examine the source data for the place attachment surfaces with a mouse click.
6.5. Implications Being able to determine the spatial distribution of place attachment yields new insights on the importance of place, not only for the entire study area, but also for subsections. For our current study area, we can combine place attachment surfaces to determine the collective place attachment for distinct classes of park users (runners, cyclists, fishers, swimmers), different demographic classes (age, ethnicity), and distinct groups by time of visitation. These types of analyses can provide new insights about how places work. Much theoretical work has been done, and representing the spatial distribution of place attachment is a logical extension of these methods. Mapping place attachment makes it more concrete, aiding place research and promoting better-informed decisions. Much remains to be done now that we have three-dimensional surfaces that show place attachment. Place attachment surfaces can serve as a foundation for future academic studies of place, and the methods and technology should be easily adaptable for use in disciplines such as planning, where place attachment plays a significant role. Acknowledgments Stuart Dixon and Chris Mueller provided invaluable help with data collection for this project. The City of Nanaimo is recognized for its open data access policy. Special thanks also go out to our advisors and reviewers who provided much advice and feedback. Funding: Vancouver Island University helped fund this project through its student work opportunity program. The university was not involved in the study design, data collection, interpretation, or analysis of data. References Augé, M. (1992). Non-Places: Introduction to and anthropology of supermodernity. London: Verso. Brown, G. (2005). Mapping spatial attributes in survey research for natural resource management: Methods and applications. Society & Natural Resources, 18, 17–39. Brown, G., & Raymond, C. (2007). The relationship between place attachment and landscape values: Toward mapping place attachment. Applied Geography, 27(2), 89–111. Brown, G., Raymond, C., & Corcoran, J. (2014). Mapping and measuring place attachment. Applied Geography, 57, 42–53. Brown, G., & Weber, D. (2011). Public participation GIS: A new method for national park planning. Landscape and Urban Planning, 102, 1–15. Burrough, P. (1996). Natural objects with indeterminate boundaries. In P. Burrough, & A. Frank (Eds.). Geographic objects with indeterminate boundaries (pp. 3–28). London: Taylor and Francis. Carver, S., Watson, A., Waters, T., Matt, R., Gunderson, K., & Davis, B. (2009). Developing computer-based participatory approaches to mapping landscape values for landscape and resource management. In S. Geertman, & J. Stillwell (Eds.). Planning support systems best practice and new methods (pp. 431–448). New York: Springer Science +Business Media. Cunningham, T. (2017). City council adds park land at Colliery dams. Nanaimo News Bulletin, 29(23), 1. Jul. 27, 2017 http://www.nanaimobulletin.com/news/citycouncil-adds-park-land-at-colliery-dams/. Dornič, S. (1967). Reports from the psychological laboratoriesSubjective distance and emotional involvement: A verification of the exponent invarianceVol. 237. Stockholm: University of Stockholm. Ekman, G., & Bratfisch, O. (1965). Subjective distance and emotional involvement: A psychological mechanism. Acta Psychologica, 24, 430–437. Evans, A., & Waters, T. (2007). Mapping vernacular geography: web-based GIS tools for capturing "fuzzy" or "vague" entities. International Journal of Technology, Policy and Management, 7(2), 134–150. Fonte, C. (2008). Geographical surfaces as entities. In W. Lodwick (Ed.). Fuzzy surfaces in GIS: Theory, analytical methods, algorithms, and applications (pp. 63–84). Boca Raton,
87
Applied Geography 99 (2018) 77–88
B. Maguire, B. Klinkenberg
Arnold. Plutchik, R. (1962). The emotions: Facts, theories, and a new model. New York: Random House. Plutchik, R. (1980). Emotion: A psychoevolutionary synthesis. New York: Harper and Row. Pred, A. (1984). Place as historically contingent Process: Structuration and the timegeography of becoming places. Annals of the AAG, 74(2), 279–297. Russell, J. (2003). Core affect and the psychological construction of emotion. Psychological Review, 110(1), 145–172. Stedman, R. (2002). Toward a social psychology of place: Predicting behavior from placebased cognitions, attitude, and identity. Environment and Behavior, 34(5), 561–581. Thrift, N. (1994). Inhuman Geographies: Landscapes of speed, light and power. In P. Cloke, M. Doel, D. Matless, M. Phillips, & N. Thrift (Eds.). Writing in the rural: Five cultural geographies (pp. 191–248). London: Paul Chapman. Vaske, J., & Kobrin, K. (2001). Place attachment and environmentally responsible behavior. The Journal of Environmental Education, 32(4), 16–21. Williams, D. (2014). Making sense of 'place': Reflections on pluralism and positionality in place research. Landscape and Urban Planning, 131, 74–82. Williams, D. (2016). Personal communication. June 27, 2016. Williams, D., & Vaske, J. (2003). The measurement of place attachment: Validity and generalisability of a psychometric approach. Forest Science, 49(6), 830–840. Zadeh, L. (1965). Fuzzy sets. Information and Control, 8, 338–353.
FL: CRC Press. Fonte, C., & Lodwick, W. (2005). Modelling the fuzzy spatial extent of geographical entities. In F. Petry, V. Robinson, & M. Cobb (Eds.). Fuzzy modelling with spatial information for geographical problems (pp. 121–142). Berlin: Springer. Golledge, R. (1978). Representing, interpreting and using cognized environments. Papers of the regional science association. Gould, P., & White, R. (1992). Mental maps (2nd ed.). New York: Routledge. Gunderson, K., & Watson, A. (2007). Understanding place meanings in the bitterroot national forest, Montana. Society & Natural Resources, 20, 705–721. Heywood, I., Cornelius, S., & Carver, S. (2011). An introduction to geographical information systems (4th ed.). Harlow, Essex: Pearson. Jorgensen, B., & Stedman, R. (2001). Sense of place as an attitude: Lakeshore owners attitudes toward their properties. Journal of Environmental Psychology, 21, 233–248. Kitchin, R., & Blades, M. (2002). The cognition of geographic space. London: I.B. Tauris. Lowery, D., & Morse, W. (2013). A qualitative method for collecting spatial data on important places for recreation, livelihoods, and ecological meanings: Integrating focus groups with public participation geographic information systems. Society & Natural Resources, 26, 1422–1437. Lynch, K. (1960). The image of the city. Cambridge, MA: MIT Press. Massey, D. (1997). A global sense of place. In T. Barnes, & D. Gregory (Eds.). Reading human geography: The poetics and politics of inquiry (pp. 315–323). London: Edwin
88